OpenAI’s Turbulent Year: Leadership Struggles and AI Milestones in 2023

Over the past 12 months, the field of artificial intelligence has witnessed numerous events—from major deals and significant updates to scandals. The technology was discussed so frequently that the British dictionary Collins named the term AI as the word of the year, while the Cambridge Dictionary chose the verb “hallucinate“.

Let’s revisit the notable developments in the realm of artificial intelligence throughout 2023.

The OpenAI Saga

Company Developments

OpenAI remains the industry’s central figure. Early 2023 was marked by a multi-billion dollar deal—Microsoft announced “multi-year” investments in the California-based AI startup. By the year’s end, it emerged that the firm was negotiating with investors about a potential stock sale, aiming to increase its valuation to $80-90 billion.

In February, the company launched a paid version of the ChatGPT chatbot, and in March, developers gained access to the neural network and the Whisper speech-to-text model via API. That same month, OpenAI introduced the large multimodal model GPT-4.

However, the company faced regulatory challenges: in April, Italian authorities ordered the blocking of ChatGPT, accusing OpenAI of “illegal personal data collection.” The management managed to resolve the issues, and the chatbot was once again accessible in the jurisdiction.

Prior to this, a technology ethics group urged the U.S. Federal Trade Commission to investigate the firm. Complaints also came from regulators in:

• France;
• Germany;
• Ireland;
• United Kingdom;
• Canada.

Amidst this, the company’s CEO, Sam Altman, addressed the U.S. Congress, urging the government to regulate AI development and usage. Subsequently, President Joe Biden’s administration issued an order establishing new AI safety and protection standards.

ChatGPT’s Evolution

In the autumn, OpenAI released a major update, enabling ChatGPT to “see, hear, and speak.” 

By this time, the chatbot could already analyze data, create Python code, generate graphs, and solve mathematical problems. The neural network even managed to scientifically debunk the “flat Earth” theory and pass a neurology exam.

In November, Altman revealed plans for developing artificial general intelligence (AGI) and details about GPT-5. Just days later, OpenAI was embroiled in perhaps the year’s most significant corporate scandal.

Dismissed—Then Reinstated

On November 17, OpenAI’s board dismissed Altman from his CEO position, allegedly due to an internal review and claims that the CEO “was not always candid” with the board.

Sources from Reuters noted that shortly before this decision, company researchers warned the board about a significant AI breakthrough that “could threaten humanity.” However, this issue did not progress further.

Most of the firm’s employees and Altman’s allies opposed the decision and threatened to resign. According to Bloomberg, Microsoft CEO Satya Nadella also expressed dissatisfaction.  

The decision to dismiss Altman lasted less than a week—by November 22, OpenAI announced his return as CEO. The entire board resigned. The new board included:

• former Salesforce co-CEO Bret Taylor (chairman);
• former U.S. Treasury Secretary Larry Summers;
• Quora co-founder Adam D’Angelo (a member of the previous board).

The AI Race

Google

In early February, Google introduced the Bard chatbot, built on the lightweight LaMDA language model. Later, the corporation launched an Experiment updates blog for it, listing changes to the dialogue AI algorithm.

In March, Bard became available to a limited number of waitlist participants in the U.S. and the UK.

In late spring, at the I/O 2023 developer conference, Google unveiled new AI features for its services. These included not only the chatbot but also the multimodal PaLM 2 model, integration of dialogue AI into Google Search, and the Duet AI toolset.

On July 13, the corporation launched Bard for users in the European Union and Brazil, introducing several new chatbot features. In December, the Gemini update was released, enhancing its performance. 

Meta

Meta, led by Mark Zuckerberg, strived to keep pace with the tech giants and industry leaders throughout the year. In February, the corporation released the large language model LLaMA for AI researchers with 13 billion and 65 billion parameters. Developers claimed that the smaller LaMMA-13B version outperformed GPT-3 from OpenAI “in most tests.”

In the third quarter, reports emerged that Meta intended to create a ChatGPT competitor. At that time, the company introduced the AI tool AudioCraft for generating music based on text descriptions.

In late September, at a presentation, the corporation showcased a voice AI assistant, neural networks with different personalities, and “smart glasses.” Two months later, Meta announced Emu Video and Emu Edit—generative AI-based tools for editing and creating content.

Meanwhile, according to media reports, the corporation decided to disband the team responsible for regulating and preventing potential threats in AI development. In December, Zuckerberg’s company and IBM announced the formation of the AI Alliance to collaborate on technology development, joined by over 50 tech firms.

Other Players

In March 2023, Microsoft introduced the Kosmos-1 neural network, integrating text, images, audio, and video content as input. Researchers described the system as a “multimodal large language model.” They believe such algorithms will form the basis of AGI, capable of performing tasks at a human level.

In the same month, search giant Baidu unveiled the AI-based Ernie Bot. The chatbot aims to revolutionize the search system and could enhance the efficiency of cloud computing, smart cars, home appliances, and other core businesses. However, the company warned that the system has several issues.

Chinese tech giant Alibaba plans to rival ChatGPT with Tongyi Qianwen. The tool supports English and Chinese languages. Initially, it will appear in the corporate messenger DingTalk. 

The chatbot will perform various tasks, including:

• converting meeting conversations into written notes;
• writing emails;
• drafting business proposals.

Elon Musk also aims to keep up with the tech giants. In November, the billionaire announced the development of the Grok neural network by his company xAI. It is modeled after the guide from Douglas Adams’ science fiction series “The Hitchhiker’s Guide to the Galaxy.” 

Following the announcement, a wave of cryptocurrencies with the same name flooded the network. Some of them turned out to be scams. 

Concerns and Risks

However, as the technology advanced, concerns about its safety emerged. Some experts expressed fears of losing control over AI development.

In March, over a thousand industry experts, including Elon Musk and Steve Wozniak, called for a six-month pause on training language models more powerful than OpenAI’s GPT-4. The letter detailed potential threats to society and civilization from competitive AI systems, including economic and political upheavals.

Nonetheless, some AI experts criticized this initiative, calling it “AI hype” and a distortion of several scientists’ articles. Microsoft founder Bill Gates expressed a similar view, stating that the world should focus on using the technology for the benefit of society.

Europol warned of threats from ChatGPT, and the European Commission saw risks to democracy in generative AI tools. 

In May, former Google CEO Eric Schmidt stated that AI poses an “existential risk,” potentially causing “harm or death” to many people. Before this, the head of OpenAI acknowledged that the technology would radically change “society as we know it,” but developers “must be cautious.”

In August, most participants in a Gartner survey identified generative AI as a threat to organizations. However, Meta deemed concerns about existential risks associated with the technology “premature.” Biden made a similar statement but emphasized that companies should be responsible for the safety of technology-related tools and products.

In December, Pope Francis warned of the dangers of a neural network dictatorship. Additionally, 2023 saw a rise in the popularity of deepfakes among fraudsters. 

Conclusions

The past year demonstrated not only the rapid development of new technologies but also the responsibility that developers must assume. Judging by the positions and statements of major industry players, they understand this.

2023 was rich in interesting and pivotal events. However, the community has much work ahead to alleviate concerns about AI and create safe and useful tools.